Research Of Distributed Clustering Algorithm Based On Density

With the rapid development of modern enterprise, data generated from different information systems become more and more. It is really not easy to extract useful information from such a vast amount of sources. How to utilize the huge original data to analyse current situation and predict future of quantities effectively, has already become a great challenge that the human beings have faced. Data Mining is devised to solve the problem.Java Data Mining framework is a standard data mining specification under Java^TM platform. In this thesis three core components of JDM and relationships among them are explained. Some interfaces and methods are discussed in detail.Cluster analysis is an important research area in data mining. Nowadays clustering has become an increasingly wide task in modern application domains such as marketing and purchasing assistance, multimedia, Biology as well as many others. In most of these areas, the data are distributed at different sites. In order to extract information from these distributed data with traditional clustering algorithm, the distributed data have to be merged at a central site and then clustered. It is such a hard topic, even incredible in some application, to collect these distributed data due to the restriction of transmission speed and safety factor.In this thesis, DBDC algorithm, which is a distributed data mining algorithm based on density, is discussed in detail.According to its deficiency in local and global clustering, an enhanced distributed data mining algorithm based on density is proposed. The algorithm mentioned above can deal with the noisy data effectively and improve the precision of distributed clustering without obvious inefficiency.